Prediction of Various Properties of Soft Ground Soils using Artificial Neural Network
Young Su Kim, Woo-Seob Jeong, Hwan Chul Jeonge, An Sik Im · Journal of the Korean Society of Civil Engineers · 2006
This study performed field and laboratory tests for poor subsoils taken in six regions of the country and determined undrain shear strength. Su values and preconsolidation pressure are predicted using Back Propagation neural network (BPNN) and the application of BPNN is verified. The result of BPNN shows that correlation coefficient between test and neural network result is over 0.9, which means high correlativity. Especially the neural network uses only 6 parameters such as natural water content, void ratio, specific gravity, rate of passing 200th sieve, liquid limits and plasticity index among various affecting factors to estimate value and the correlation coefficent is 0.93. The conclusions obtained in this paper are from the tests performed for poor subsoils taken in the several regions of the country. If there were more test results, the prediction and influence of various soil properties could be effectively performed by neural network.